I have a BGP free core (MPLS) and a number of PE devices all connected to RR's. Basic diagram here
I have a customer with their own AS number connected (point to point) to a PE but advertising and receiving full routing table via eBGP multihop to a RR.
Routes are received and advertised fine and has been working fine as well.
For example the RR receives the customer owned prefix 8.8.8.0/24 via the eBGP multihop and advertises it out. The prefix is then also reachable as traffic to 8.8.8.0/24 will go via the RR which has the next-hop as the point to point circuit and all is good.
I've just swapped out a RR from IOS-XE to a IOS-XR unit and everything works fine except for this ebgp multihop peering.
We again receive the route fine on the RR and the customer is receiving the full routing table but when you try to route to 8.8.8.0/24 it will take the same path but on reaching the PE it will see the route as being on the RR and it will loop back and forth. For some reason the route is not having it's next hop changed to the remote end of the PE.
I've tested this myself and I've made sure the next-hop-self is set on the eBGP peering on the CE and it doesn't make a difference.
If I change the eBGP peering so it instead uses a RR with IOS-XE setup it works fine. I've compared configs and cannot see anything missed.
Test CE router is simply using ebgp multihop and next-hop-self
RR is setup with same ebgp multihop.
Tried with next-hop-self enabled and disabled. also tried enabling 'next-hop-self inheritance-disable'.
No comments:
Post a Comment